Metal Oxide Semiconductor Field-Effect Transistors (MOSFETs), also known as Insulated Gate Field-Effect Transistors (IGFETs), are semiconductor devices in which carriers (electrons and/or holes) are provided by one or more source(s), traverse one or more channel(s), and are collected by one or more drain(s). Resistance of the channel(s) is then controlled by one or more gate(s). For high performance, the resistance associated with the electrical path between source(s) and drain(s) should be low when the transistor channel(s) are “on”, and high when the transistor is “off”. Source(s) and/or drain(s) are typically formed of relatively highly conducting regions to minimize the electrical resistance of the device when channel(s) are on.
A key to a low resistance between source(s) and drain(s) is that the resistance between source(s) and channel(s) be low when the channel(s) are on. This requires a sufficiently unimpeded current path between the highly conducting source(s) and the channel(s). The resistance between drain(s) and channel(s) can also be important. Among the ways in which the current path between a source (or drain) and a channel can be impeded is a potential barrier between the source (or drain) and the channel. Exemplary band diagrams from a metal source, through a silicon channel, and into a metal drain are shown in FIGS. 1(a) and 1(b).
In particular, FIG. 1(a) illustrates an example with a zero potential barrier between a source 10 and a channel 12 and between a drain 14 and the channel 12. Carriers flow efficiently from the source into the channel, from which they are subsequently collected by the drain. FIG. 1(b) illustrates an example of a potential barrier between a source 10′ (and drain 14′) and a channel 12′. Carriers must be emitted over, or tunnel through, the potential barrier at the source side to reach the channel from the source. This reduces the available device current relative to the case of a zero potential barrier.
One class of field-effect transistors is “metal source/drain” transistors. In these devices, the source(s) and drain(s) consist of metals, rather than the impurity-rich semiconductors commonly employed in the art. The advantage of metal source/drain FETs, over FETs with sources and drains consisting of impurity-doped semiconductor(s), is that metals generally have higher conductivity than do impurity-doped semiconductors, allowing for the possibility of a lower resistance transistor. An additional advantage is that metal source(s) and drain(s) can be made to be atomically abrupt, whereas impurity profiles in semiconductors tend to have more gradual transitions between impurity-rich and impurity-poor regions, or between regions rich in donor impurities and regions rich in acceptor impurities. Derivatives of metal source/drain transistors include metal source transistors, in which source(s) but possibly not drain(s) are metal, and metal drain transistors, in which drain(s) but possibly not source(s), are metal.
In such devices, a critical parameter is the effective workfunction of the metal at its interface with the semiconductor(s) forming the region of the channel(s). For example, for an n-channel FET, in which the channel is formed primarily of electrons in the semiconductor, a low effective workfunction associated with this interface may be preferred, to minimize or eliminate a potential barrier from the metal to the channel. Alternatively, for a p-channel FET, in which the channel is formed primarily of holes in the semiconductor, a high effective workfunction associated with this interface may be preferred, to minimize or eliminate a potential barrier from the metal to the channel.
One approach to forming metal sources and/or drains is via reacting a metal with one or more semiconductors in the vicinity of the source and/or drain regions to form metal-semiconductor compounds. For example, with Si-based transistors (Si making up the material in which the channel is formed), metal silicides are used. See, for example U.S. Pat. No. 6,744,103 of Snyder, “Short-channel Schottky-Barrier MOSFET Device and Manufacturing Method”. 
P-channel and n-channel transistors can be fabricated in the same process with the use of two silicides, one with a relatively low workfunction for n-channel transistors, and the other with a relatively high workfunction for p-channel transistors. For example, Kedzierski et al., “Complementary silicide source/drain thin-body MOSFETs for 20 nm gate length regime”, 2000 IEEE IEDM Technical Digest, pp. 57-60, use ErSi1.7 as the low-workfunction silicide, and PtSi as the high-workfunction silicide. The process described is, in part: (a) deposit Pt, (b) pattern the Pt, (c) apply heat to cause PtSi to be formed, (d) remove the unreacted Pt, (e) deposit Er, (f) pattern the Er, (g) apply heat to form ErSi1.7, (h) remove the unreacted Er.
Another example is provided by X. Tang, et al., “Very low Schottky barrier to n-type silicon with PtEr-stack silicide”, Solid-State Electronics, v. 47, pp 2105-2111 (2003). They describe a process in which 50 nm of Er is capped with 40 nm of Pt, and then exposed to heat to form ErSi1.7, while the Pt remains unreacted. The motivation for the process is to use the Pt for the formation of PtSi on, for example, p-channel FETs, while ErSi1.7 is formed on, for example, n-channel FETs. Further, it is claimed the quality of the ErSi1.7 is improved by capping the Er during the silicidation reaction. A full process is not described, nor were transistors fabricated, and so whether the unreacted Pt remains in place at the completion of the process is not disclosed. No motivation was presented by the authors for doing so, and doing so would be contrary to the present art. Therefore, there is no indication the work was a preparation for a multi-layer source/drain, and certainly the work was not a preparation for a deposited source/drain consistent with the use of the term “deposited source/drain” in the context of the present invention.
It has also been established (see S. Park, et al., “Ab initio study of metal gate electrode work function,” J. Appl. Phys, vol. 86, p. 073118 (2005)), that the effective workfunction of metals at interfaces is dominated by the metal within the first few atomic layers of the interface. Therefore, for metal source(s) and/or metal drain(s), more than one metal may be present, but the metal within a few atomic layers of the interface is the one which primarily determines the existence and/or extent of a potential barrier between the source(s) and/or drain(s) and channel(s) with which they are in contact.